[FEATURE] script_score sub-type in rerank processor

[joelthe1]

Is your feature request related to a problem?

I have successfully setup and run hybrid search using OpenSearch on my data but I want to tune the results so that the results which are larger in (string) length (for a specific field) are boosted slightly over those results which are relatively smaller in length. The motivation for this is that sometimes, semantic search returns results that are much shorter (think a sentence) than a very long query string (think a paragraph of text) and I would like to reduce the relevance of such results.

What solution would you like?

Ability to write my own simple script to rerank like in the script_score function.

What alternatives have you considered?

An inferior alternative would be to use script_score (within function_score ) and compare the length of the query string with the length of the result. But this comparison is different from the request, since it is a comparison between the individual search result and the query. What I want is to compare all the search results with each other using some simple function (like script_score) which is not currently possible.

Do you have any additional context?

Using the Painless scripting language might be more fitting.

[navneet1v]

@joelthe1 if I understand the feature request properly, you are requesting an ability in re-ranking processor which can run a custom logic on the final search results to re-ranking the documents? is that a correct understanding?

[vamshin]

@HenryL27 will you be able to take look?

[HenryL27]

I'm not gonna have time for this, no. Sorry

[vamshin]

np. @joelthe1 do you think you can help adding this feature?

[joelthe1]

@joelthe1 if I understand the feature request properly, you are requesting an ability in re-ranking processor which can run a custom logic on the final search results to re-ranking the documents? is that a correct understanding?

Yes @navneet1v, that is correct. Unfortunately, I am not in a position to work on this myself but do think it would benefit more than just me. However, I am able to test things or provide feedback, if useful.

[heemin32]

@joelthe1 Could you provide more details to clarify the request? For example, an API example and the expected behavior would be helpful.